I think it is a basic feature that every language should have.
Right now, every time Alexa doesn't understand something it randomly select an intent and answer something unexpected.
For example when I ask something related to my app that is still not trained it will greet the user. This doesn't make sense at all. If the confidence of the answer is too low it should be managed in that "fallback" intent.
Meanwhile is there any way to check the confidence of the answer in the code so I can manually be able to manage the answer?